I do think Jane Street fell into a sunk cost fallacy when they first stuck to OCaml. But after having invested further into an OCaml/OxCaml codebase, teams and tooling, switching tech stacks has lost most of its value.
Also, their efforts benefit me greatly as an OCaml (now maybe OxCaml) user, so I'm glad they took that path.
I dunno. Jane Street started using OCaml back in 2002. Consider the world of programming languages back then: no Rust, of course...but, like Java wouldn't even have generics for another two years! At the time - and for a long time after - OCaml's offering of power/performance was pretty unbeatable. It arguably still is, but I think there's at least an argument these days. I'm not sure exactly when that argument became viable, but it would've been long-after Jane Street had already built up a pretty huge codebase and a pretty crack team of OCaml engineers.
Also, their efforts benefit me greatly as an OCaml (now maybe OxCaml) user, so I'm glad they took that path.